Objective: To investigate the accuracy of three clinical scales for predicting severe disease (severe dehydration or death) in children with diarrhea in a resource-limited setting. Methods: Participants included 178 children admitted to three Rwandan hospitals with diarrhea. A local physician or nurse assessed each child on arrival using the World Health Organization (WHO) severe dehydration scale and the Centers for Disease Control (CDC) scale. Children were weighed on arrival and daily until they achieved a stable weight, with a 10% increase between admission weight and stable weight considered severe dehydration. The Clinical Dehydration Scale was then constructed post-hoc using the data collected for the other two scales. Receiver Operator Characteristic (ROC) curves were constructed for each scale compared to the composite outcome of severe dehydration or death. Results: The WHO severe dehydration scale, CDC scale, and Clinical Dehydration Scale had areas under the ROC curves (AUCs) of 0.72 (95% CI 0.60, 0.85), 0.73 (95% CI 0.62, 0.84), and 0.80 (95% CI 0.71, 0.89), respectively, in the full cohort. Only the Clinical Dehydration Scale was a significant predictor of severe disease when used in infants, with an AUC of 0.77 (95% CI 0.61, 0.93), and when used by nurses, with an AUC of 0.78 (95% CI 0.63, 0.93). Conclusions: While all three scales were moderate predictors of severe disease in children with diarrhea, scale accuracy varied based on provider training and age of the child. Future research should focus on developing or validating clinical tools that can be used accurately by nurses and other less-skilled providers to assess all children with diarrhea in resource-limited settings.